Overview

One-way ANOVA tests whether the means of three or more groups differ. This takes the sums of squares and degrees of freedom from an ANOVA table and works out the mean squares and the F statistic.

Enter the between-group and within-group sums of squares and degrees of freedom, and your alpha. You get the exact p-value from the F distribution.

Worked examples

InputsResult
SS Between460DF Between3SS Within980DF Within44Alpha (%)5MS between: 153.33MS within: 22.27F-statistic: 6.88p-value: 0.000668Significant at alpha=5%
Show 2 more examples
SS Between120DF Between2SS Within600DF Within27Alpha (%)5MS between: 60MS within: 22.22F-statistic: 2.7p-value: 0.09Not significant at alpha=5%
SS Between50DF Between4SS Within1,000DF Within95Alpha (%)5MS between: 12.5MS within: 10.53F-statistic: 1.19p-value: 0.32Not significant at alpha=5%

MS between by DF between

With the calculator's starting values, the MS between is 153.33. The table shows how it changes with the DF between.

DF BetweenMS between
1.5306.67
2230
2.5184
3starting value153.33
3.5131.43
4115
4.5102.22

Other inputs: SS Between: 460 · SS Within: 980 · DF Within: 44 · Alpha: 5%. Change them in the calculator above for your own numbers.

How it works

Each sum of squares divided by its degrees of freedom gives a mean square. F is the between-group mean square divided by the within-group one. An F near 1 suggests the groups don't differ, while a large F suggests at least one mean is different. The p-value comes from the F distribution with both degrees of freedom.

The formula

F = (SS_between ÷ df_between) ÷ (SS_within ÷ df_within)
df_betweenNumber of groups − 1
df_withinTotal observations − number of groups

Practical tips

  • SS between 460 on 3 df and SS within 980 on 44 df give F = 6.88. The 5% critical value is about 2.82, so it's significant.
  • You can check the p-value in Excel with =F.DIST.RT(6.88, 3, 44).
  • A significant F only says some means differ. Use a post-hoc test such as Tukey's to find which.

Frequently asked questions

What are the assumptions?

Independent observations, roughly normal data in each group, and similar spread across groups.

Why not run several t-tests?

Each extra test raises the chance of a false positive. ANOVA tests all groups at once.

How is the p-value worked out?

From the F distribution with your two degrees of freedom. It's the chance of an F this large if the group means were really equal.